The HP-Epson comparison bore out the one experiment I'd participated in. And it led me to a speculation: The HP printer may have more potential than its convenient profiling system lets you see. This system has the wonderful advantage of ease of use, but it looks like there’s evidence that it employs fewer than the optimal array of color samples.
In the Epson-HP-Canon comparison that I’d seen, the printers were an iPF5000 with a profile by Andrew Rodney, a 9800 with ImagePrint, and an HP set up by its distributor (not, presumably, with the new firmware). The most important output was an 8-bit target from HP. This had been printed on all 3 printers, using HPR on the Epson and Canon, and HP's version of rag on the Z3100. (The 8-bit target didn't allow the Canon to print in its high-bit mode).
On the score of color reproduction we observed the same things as in the report, especially re: blues. But we noticed especially that the HP had trouble with chroma, that is, with the differentiation of hues from gray as you move up in lightness from black. Or in simpler words, it's about where color appears in shadow detail. It shows up test prints that represent a spectrum on the X axis & go from black to white on the Y axis. Uneven emergence of hues from black was noticeable in comparing the HP’s output to the other two.
This weakness isn’t necessarily a problem with a printer; it can also appear when using profiles from a limited number of swatches. (Remember the early Colorvision system?) The review noticed that the HP had more trouble differentiating the lower tones of a gray scale. Add color and you have relatively more 'murk' before clear colors appear in the shadows. This is where the report says some further investigation would be a good idea.
This leaves me wondering if the HP profiling system, easy to operate and desirable in conjunction with inexpensive replaceable heads, may be holding back the Z3100's performance in printer-to-printer comparisons. The regular HP spectrophotometer measures only a few hundred patches, and the advanced version, if I understand correctly, doesn't quite make it to 1000. But you can see chroma differences in test prints from profiles made from 1700 swatch & 4000 swatch targets. This is basically the difference between good though inexpensive profiles of small file size (for example from Inkjetart) and top-quality profiles with larger file size (for example profiles made by Andrew Rodney from Bill Atkinson’s larger set of targets).
P.S.: While preferring the Canon for slight but observable differences, I agree with the review and with Michael that HP, Epson, and Canon are all batting in the same league from the standpoint of image quality. Differences in IQ are small enough that they may not matter as much as reliability, customer service, or relationship with a dealer - and price and size. The HP advantage is that it reduces gloss differential on some, though not all, papers (satin/luster, better than glossy). And Canon takes a hit on warranty policies and customer service (see Wiki). My suspicion is that the ‘world printer series’ of the moment goes to HP, especially if you use profiles from the large Atkinson targets for your finest work.